Original commit message from CVS:
* ext/faad/gstfaad.c: (gst_faad_open_decoder):
FAAD fails to decode low (e.g. 8 kHz) sample rate AAC data in
quicktime because of sample rate mismatches.
Reenable overriding the implicit SBR behaviour (accidently changed?)
to allow playback of these files.
Original commit message from CVS:
* configure.ac:
Disable rtpmanager for now because it depends on CVS -base.
* gst/rtpmanager/Makefile.am:
Added new files for session manager.
* gst/rtpmanager/gstrtpjitterbuffer.h:
* gst/rtpmanager/gstrtpbin.c: (create_session), (get_pt_map),
(create_stream), (pt_map_requested), (new_ssrc_pad_found):
Some cleanups.
the session manager can now also request a pt-map.
* gst/rtpmanager/gstrtpsession.c: (gst_rtp_session_base_init),
(gst_rtp_session_class_init), (gst_rtp_session_init),
(gst_rtp_session_finalize), (rtcp_thread), (start_rtcp_thread),
(stop_rtcp_thread), (gst_rtp_session_change_state),
(gst_rtp_session_process_rtp), (gst_rtp_session_send_rtp),
(gst_rtp_session_send_rtcp), (gst_rtp_session_clock_rate),
(gst_rtp_session_get_time), (gst_rtp_session_event_recv_rtp_sink),
(gst_rtp_session_chain_recv_rtp),
(gst_rtp_session_event_recv_rtcp_sink),
(gst_rtp_session_chain_recv_rtcp),
(gst_rtp_session_event_send_rtp_sink),
(gst_rtp_session_chain_send_rtp), (create_send_rtcp_src),
(gst_rtp_session_request_new_pad):
* gst/rtpmanager/gstrtpsession.h:
We can ask for pt-map now too when the session manager needs it.
Hook up to the new session manager, implement the needed callbacks for
pushing data, getting clock time and requesting clock-rates.
Rename rtcp_src to send_rtcp_src to make it clear that this RTCP is to
be send to clients.
Add code to start and stop the thread that will schedule RTCP through
the session manager.
* gst/rtpmanager/rtpsession.c: (rtp_session_class_init),
(rtp_session_init), (rtp_session_finalize),
(rtp_session_set_property), (rtp_session_get_property),
(on_new_ssrc), (on_ssrc_collision), (on_ssrc_validated),
(on_bye_ssrc), (rtp_session_new), (rtp_session_set_callbacks),
(rtp_session_set_bandwidth), (rtp_session_get_bandwidth),
(rtp_session_set_rtcp_bandwidth), (rtp_session_get_rtcp_bandwidth),
(source_push_rtp), (source_clock_rate), (check_collision),
(obtain_source), (rtp_session_add_source),
(rtp_session_get_num_sources),
(rtp_session_get_num_active_sources),
(rtp_session_get_source_by_ssrc),
(rtp_session_get_source_by_cname), (rtp_session_create_source),
(update_arrival_stats), (rtp_session_process_rtp),
(rtp_session_process_sr), (rtp_session_process_rr),
(rtp_session_process_sdes), (rtp_session_process_bye),
(rtp_session_process_app), (rtp_session_process_rtcp),
(rtp_session_send_rtp), (rtp_session_get_rtcp_interval),
(rtp_session_produce_rtcp):
* gst/rtpmanager/rtpsession.h:
The advanced beginnings of the main session manager that handles the
participant database of RTPSources, SSRC probation, SSRC collisions,
parse RTCP to update source stats. etc..
* gst/rtpmanager/rtpsource.c: (rtp_source_class_init),
(rtp_source_init), (rtp_source_finalize), (rtp_source_new),
(rtp_source_set_callbacks), (rtp_source_set_as_csrc),
(rtp_source_set_rtp_from), (rtp_source_set_rtcp_from),
(push_packet), (get_clock_rate), (calculate_jitter),
(rtp_source_process_rtp), (rtp_source_process_bye),
(rtp_source_send_rtp), (rtp_source_process_sr),
(rtp_source_process_rb):
* gst/rtpmanager/rtpsource.h:
Object that encapsulates an SSRC and its state in the database.
Calculates the jitter and transit times of data packets.
* gst/rtpmanager/rtpstats.c: (rtp_stats_init_defaults),
(rtp_stats_calculate_rtcp_interval), (rtp_stats_add_rtcp_jitter):
* gst/rtpmanager/rtpstats.h:
Various stats regarding the session and sources.
Used to calculate the RTCP interval.
Original commit message from CVS:
* gst/nsf/types.h:
Rename #ifndef header guard symbol to something less generic, so
types.h doesn't get skipped over when compiling on MingW. Include
GLib headers and use those to set the endianness and the basic
types so that this isn't entirely broken for non-x86 architectures.
Original commit message from CVS:
* gst/mve/gstmvedemux.c: (gst_mve_audio_init):
Use G_LITTLE_ENDIAN instead of LITTLE_ENDIAN, so stuff compiles on
MingW (no idea though why we add a BYTE_ORDER endianness field if
the audio is compressed).
Original commit message from CVS:
* gst/rtpmanager/gstrtpbin-marshal.list:
Some more custom marshallers.
* gst/rtpmanager/gstrtpbin.c: (create_session), (get_pt_map),
(clock_rate_request), (create_stream), (gst_rtp_bin_class_init),
(pt_map_requested), (new_ssrc_pad_found), (create_recv_rtp):
* gst/rtpmanager/gstrtpbin.h:
Prepare for caching pt maps.
Connect to signals to collect pt maps.
* gst/rtpmanager/gstrtpjitterbuffer.c:
(gst_rtp_jitter_buffer_class_init),
(gst_jitter_buffer_sink_setcaps), (gst_rtp_jitter_buffer_loop):
* gst/rtpmanager/gstrtpjitterbuffer.h:
Add request_clock_rate signal.
Use scale insteat of scale_int because the later does not deal with
negative numbers.
* gst/rtpmanager/gstrtpptdemux.c: (gst_rtp_pt_demux_class_init),
(gst_rtp_pt_demux_chain):
* gst/rtpmanager/gstrtpptdemux.h:
Implement request-pt-map signal.
Original commit message from CVS:
* gst/rtpmanager/gstrtpbin.c: (gst_rtp_bin_class_init),
(gst_rtp_bin_init), (gst_rtp_bin_provide_clock):
* gst/rtpmanager/gstrtpbin.h:
Provide a clock.
Original commit message from CVS:
* gst/rtpmanager/gstrtpjitterbuffer.c:
(gst_jitter_buffer_sink_setcaps), (gst_rtp_jitter_buffer_chain),
(gst_rtp_jitter_buffer_loop):
Add some debug and comments.
Fix double unref() in error cases.
Original commit message from CVS:
* gst/qtdemux/gstrtpxqtdepay.c: (gst_rtp_xqt_depay_process),
(gst_rtp_xqt_depay_change_state):
* gst/qtdemux/gstrtpxqtdepay.h:
Try to recover from packet loss a little better.
Original commit message from CVS:
* ext/jack/gstjackaudiosink.c: (gst_jack_ring_buffer_open_device),
(gst_jack_ring_buffer_acquire):
Try t better name clients. properly handle return codes when re-
establishing links.
Original commit message from CVS:
* ext/wavpack/gstwavpackdec.c: (gst_wavpack_dec_reset),
(gst_wavpack_dec_init), (gst_wavpack_dec_sink_set_caps),
(gst_wavpack_dec_clip_outgoing_buffer),
(gst_wavpack_dec_post_tags), (gst_wavpack_dec_chain):
* ext/wavpack/gstwavpackdec.h:
* ext/wavpack/gstwavpackenc.c: (gst_wavpack_enc_reset),
(gst_wavpack_enc_sink_set_caps), (gst_wavpack_enc_set_wp_config),
(gst_wavpack_enc_chain):
* ext/wavpack/gstwavpackenc.h:
* ext/wavpack/gstwavpackparse.c:
Don't play audioconvert. As wavpack wants/outputs all samples with
width==32 and depth=[1,32] accept this and let audioconvert convert
to accepted formats instead of doing it in the element for n*8 depths.
This also adds support for non-n*8 depths and prevents some useless
memory allocations. Fixes#421598
Also add a workaround for bug #421542 in wavpackenc for now...
* tests/check/elements/wavpackdec.c: (GST_START_TEST):
* tests/check/elements/wavpackenc.c: (GST_START_TEST):
* tests/check/elements/wavpackparse.c: (GST_START_TEST):
Consider the change above in the unit tests and test if the correct
caps are accepted and set. Also check for GST_BUFFER_OFFSET_END in
the wavpackparse unit test.
* ext/wavpack/gstwavpackdec.c: (gst_wavpack_dec_init),
(gst_wavpack_dec_sink_set_caps):
Set caps on the src pad as soon as possible.
* ext/wavpack/gstwavpackdec.h:
* ext/wavpack/gstwavpackcommon.h:
* ext/wavpack/gstwavpackenc.h:
* ext/wavpack/gstwavpackparse.h:
Fix indention. gst-indent is now called by cicl.
Original commit message from CVS:
* gst/qtdemux/qtdemux.c: (gst_qtdemux_prepare_current_sample),
(gst_qtdemux_chain), (qtdemux_parse_samples):
* gst/qtdemux/qtdemux_dump.c: (qtdemux_dump_ctts):
* gst/qtdemux/qtdemux_dump.h:
* gst/qtdemux/qtdemux_fourcc.h:
* gst/qtdemux/qtdemux_types.c:
Process 'ctts' atoms, which are present in AVC ISO files (.mov files
with h264 video).
Use the offset present in 'ctts' to calculate the PTS for each packet
and set the PTS on outgoing buffers.
Fixes#423283
Original commit message from CVS:
2007-03-27 Julien MOUTTE <julien@moutte.net>
* ext/xvid/gstxviddec.c: (gst_xviddec_chain): Add some
debug log and fix a stupid output buffer duration bug.
Original commit message from CVS:
* gst/interleave/deinterleave.c: (gst_deinterleave_sink_setcaps):
Remove 'channel-positions' field when munging input caps into
1-channel output caps (I guess technically we should set the
position for each channel on the output caps if it's non-NONE,
but I'll save that as a task for another day).
Original commit message from CVS:
* gst/vmnc/vmncdec.c: (gst_vmnc_dec_class_init),
(gst_vmnc_dec_init), (vmnc_dec_finalize), (gst_vmnc_dec_reset),
(vmnc_handle_wmvi_rectangle), (render_colour_cursor),
(render_cursor), (vmnc_make_buffer), (vmnc_handle_wmvd_rectangle),
(vmnc_handle_wmve_rectangle), (vmnc_handle_wmvf_rectangle),
(vmnc_handle_wmvg_rectangle), (vmnc_handle_wmvh_rectangle),
(vmnc_handle_wmvj_rectangle), (render_raw_tile), (render_subrect),
(vmnc_handle_raw_rectangle), (vmnc_handle_copy_rectangle),
(vmnc_handle_hextile_rectangle), (vmnc_handle_packet),
(vmnc_dec_setcaps), (vmnc_dec_chain_frame), (vmnc_dec_chain),
(vmnc_dec_set_property), (vmnc_dec_get_property):
Redesign to include a parser for raw files (no timestamps in that
mode yet, though).
Original commit message from CVS:
* gst/interleave/deinterleave.c: (gst_deinterleave_add_new_pads),
(gst_deinterleave_remove_pads), (gst_deinterleave_process),
(gst_deinterleave_chain):
Don't leak input buffer in chain function; maintain our own list of
source pads - there are no guarantees about the order of the list
in the GstElement struct, and we want a very specific order; lastly,
some more debugging.
Original commit message from CVS:
* ext/wavpack/gstwavpackparse.c: (gst_wavpack_parse_plugin_init):
Revert last commit, preventing infinite plugging loops with ranks
is no clean solution and in general there's no reason why one wants
to parse framed wavpack data again.
Original commit message from CVS:
* ext/wavpack/gstwavpackenc.c: (gst_wavpack_enc_push_block):
Send the new segment event in time format instead of bytes. This
allows "wavpackenc ! wavpackdec ! someaudiosink" pipelines.
* ext/wavpack/gstwavpackparse.c: (gst_wavpack_parse_plugin_init):
Accept framed and non-framed input, wavpackparse doesn't care. To
prevent "wavpackparse ! wavpackparse ! ..." pipelines lower the
rank of wavpackparse by one. This allows "wavpackenc ! wavpackparse !
..." pipelines.
Original commit message from CVS:
* gst-libs/gst/app/Makefile.am:
Use GST_ALL_LDFLAGS, which actually exists, but maybe David
can confirm that was what he wanted.
Original commit message from CVS:
* ext/wavpack/gstwavpackdec.c: (gst_wavpack_dec_chain):
Revert to use gst_pad_alloc_buffer() here. We can and should use it.
Thanks to Jan and Mike for noticing my mistake.
Original commit message from CVS:
* ext/wavpack/gstwavpackenc.c: (gst_wavpack_enc_class_init),
(gst_wavpack_enc_init), (gst_wavpack_enc_chain),
(gst_wavpack_enc_rewrite_first_block):
* ext/wavpack/gstwavpackenc.h:
Put the write helpers into the GstWavpackEnc struct directly and not
as a pointer to save two small, but useless mallocs. This also makes
it possible to drop the finalize method.
* ext/wavpack/gstwavpackparse.c: (gst_wavpack_parse_push_buffer):
For consistency reasons also set GST_BUFFER_OFFSET_END on the outgoing
buffers the same way wavpackenc does it.
Original commit message from CVS:
* ext/wavpack/gstwavpackdec.c: (gst_wavpack_dec_chain):
Don't use gst_pad_alloc_buffer() as we might clip the buffer later and
BaseTransform-based elements will likely break because of wrong
unit-size. Also plug a possible memleak that happens when decoding
fails for some reason.
Original commit message from CVS:
Based on patch by: Paul Davis <paul at linuxaudiosystems dot com>
* ext/jack/gstjackaudioclient.c: (gst_jack_audio_unref_connection):
Don't need to take the connection lock, it will not be used and could
cause deadlocks.
Original commit message from CVS:
* sys/osxvideo/osxvideosink.m:
Emit 'have-ns-view' message when working in embedded mode. The message
will contain a pointer to the newly created NSView.
Original commit message from CVS:
* gst/mpegvideoparse/mpegpacketiser.c: (mpeg_util_find_start_code),
(collect_packets), (set_par_from_dar), (set_fps_from_code),
(mpeg_util_parse_extension_packet), (mpeg_util_parse_sequence_hdr),
(mpeg_util_parse_picture_hdr):
* gst/mpegvideoparse/mpegpacketiser.h:
* gst/mpegvideoparse/mpegvideoparse.c:
(mpegvideoparse_handle_sequence), (mpegvideoparse_handle_picture),
(mpegvideoparse_drain_avail), (gst_mpegvideoparse_chain),
(mpv_parse_sink_event), (plugin_init):
* gst/mpegvideoparse/mpegvideoparse.h:
Move the MPEG specific byte parsing into the mpegpacketiser code.
Add parsing of picture types, that just feeds into a debug message
for now.
Fix some 64-bit format strings.
Original commit message from CVS:
* gst/equalizer/gstiirequalizer10bands.c:
(gst_iir_equalizer_10bands_init):
A 10 band EQ should be initialized to 1 bands and not to 3.
Original commit message from CVS:
* configure.ac:
* gst/mpeg1videoparse/Makefile.am:
* gst/mpeg1videoparse/gstmp1videoparse.c:
* gst/mpeg1videoparse/gstmp1videoparse.h:
* gst/mpeg1videoparse/mp1videoparse.vcproj:
* gst/mpegvideoparse/Makefile.am:
* gst/mpegvideoparse/mpegpacketiser.c: (mpeg_packetiser_init),
(mpeg_packetiser_free), (mpeg_packetiser_add_buf),
(mpeg_packetiser_flush), (mpeg_find_start_code),
(get_next_free_block), (complete_current_block),
(append_to_current_block), (start_new_block), (handle_packet),
(collect_packets), (mpeg_packetiser_handle_eos),
(mpeg_packetiser_get_block), (mpeg_packetiser_next_block):
* gst/mpegvideoparse/mpegpacketiser.h:
* gst/mpegvideoparse/mpegvideoparse.c: (mpegvideoparse_get_type),
(gst_mpegvideoparse_base_init), (gst_mpegvideoparse_class_init),
(mpv_parse_reset), (gst_mpegvideoparse_init),
(gst_mpegvideoparse_dispose), (set_par_from_dar),
(set_fps_from_code), (mpegvideoparse_parse_seq),
(gst_mpegvideoparse_time_code), (gst_mpegvideoparse_flush),
(mpegvideoparse_drain_avail), (gst_mpegvideoparse_chain),
(mpv_parse_sink_event), (gst_mpegvideoparse_change_state),
(plugin_init):
* gst/mpegvideoparse/mpegvideoparse.h:
* gst/mpegvideoparse/mpegvideoparse.vcproj:
Port mpeg1videoparse to 0.10 and give it rank SECONDARY-1, so
that it's below existing decoders.
Rename it to mpegvideoparse to reflect that it handles MPEG-1 and
MPEG-2 now.
Re-write the parsing code so that it collects packets differently
and timestamps Picture packets correctly.
Add a list of FIXME's at the top.
Original commit message from CVS:
* tests/icles/equalizer-test.c: (equalizer_set_band_value),
(equalizer_set_all_band_values),
(equalizer_set_band_value_and_wait),
(equalizer_set_all_band_values_and_wait), (do_slider_fiddling),
(main):
Port the example to new equalizer api.
Original commit message from CVS:
* sys/osxvideo/cocoawindow.h:
* sys/osxvideo/cocoawindow.m:
* sys/osxvideo/osxvideosink.h:
* sys/osxvideo/osxvideosink.m:
Fix leaks when running a NSApp.
Accept any kind of resolutions.
Works in fullscreen. Can maximize.
Only thing left before being able to move this to -good is documentation
and embedded window support.
